3D Model based Object Class Detection in An Arbitrary View

Size: px
Start display at page:

Download "3D Model based Object Class Detection in An Arbitrary View"

Transcription

1 3D Model based Object Class Detection in An Arbitrary View Pingkun Yan, Saad M. Khan, Mubarak Shah School of Electrical Engineering and Computer Science University of Central Florida vision Abstract In this paper, a novel object class detection method based on 3D object modeling is presented. Instead of using a complicated mechanism for relating multiple 2D training views, the proposed method establishes spatial connections between these views by mapping them directly to the surface of 3D model. The 3D shape of an object is reconstructed by using a homographic framework from a set of model views around the object and is represented by a volume consisting of binary slices. Features are computed in each 2D model view and mapped to the 3D shape model using the same homographic framework. To generalize the model for object class detection, features from supplemental views are also considered. A codebook is constructed from all of these features and then a 3D feature model is built. Given a 2D test image, correspondences between the 3D feature model and the testing view are identified by matching the detected features. Based on the 3D locations of the corresponding features, several hypotheses of viewing planes can be made. The one with the highest confidence is then used to detect the object using feature location matching. Performance of the proposed method has been evaluated by using the PASCAL VOC challenge dataset and promising results are demonstrated. 1. Introduction In recent years, the problem of object detection has received considerable attention from both the computer vision and machine learning communities. The key challenge of this problem is the ability to recognize any member in a category of objects in spite of wide variations in visual appearance due to geometrical transformations, change in viewpoint, or illumination. In this paper, a novel 3D feature model based object class detection method is proposed to deal with these challenges. The objective of this work is to detect the object given an arbitrary 2D view using a general 3D feature model of the class. In our work, the objects can be arbitrarily transformed (with translation and rotation), and the viewing position and orientation of the camera is arbitrary as well. In addition, camera parameters are assumed to be unknown. Object detection in such a setting has been considered a very challenging problem due to various difficulties of geometrically modeling relevant 3D object shapes and the effects of perspective projection. In this paper, we exploit a recently proposed 3D reconstruction method using homographic framework for 3D object shape reconstruction. Given a set of 2D images of an object taken from different viewpoints around the object with unknown camera parameters, which are called model views, the 3D shape of this specific object can be reconstructed using the homographic framework proposed in [10]. In our work, 3D shape is represented by a volume consisting of binary slices with 1 denoting the object and 0 for background. By using this method, we can not only reconstruct 3D shapes for the objects to be detected, but also have access to the homographies between the 2D views and the 3D models, which are then used to build the 3D feature model for object class detection. In the feature modeling phase of the proposed method, SIFT features [12] are computed for each of the 2D model views and mapped to the surface of the 3D model. Since it is difficult to accurately relate 2D coordinates to a 3D model by projecting the 3D model to a 2D view (with unknown camera parameters), we propose to use a homography transformation based algorithm. Since the homographies have been obtained during the 3D shape reconstruction process, the projection of a 3D model can be easily computed by integrating the transformations of slices from the model to a particular view, as opposed to directly projecting the entire model by estimation of the projection matrix. To generalize the model for object class detection, images of other objects of the class are used as supplemental views. Features from these views are mapped to the 3D model in the same way as for those model views. A codebook is constructed from all of these features and then a 3D feature model is built. The 3D feature model thus combines the 3D shape information and appearance features for robust object class detection.

2 Given a new 2D test image, correspondences between the 3D feature model and this testing view are identified by matching feature. Based on the 3D locations of the corresponding features, several hypotheses of viewing planes can be made. For each hypothesis, the feature points are projected to the viewing plane and aligned with the features in the 2D testing view. A confidence is assigned to each hypothesis and the one with the highest confidence is then used to produce the object detection result. The rest of the paper is organized as follows. Section 2 provides a brief review of related work. A summary of the 3D shape model construction method is presented in Section 3. The 3D feature model and our object class detection approach are presented in Sections 4 and 5, respectively. Section 6 provides the detection results, and finally, Section 7 concludes the work. 2. Related Work As the approaches for recognizing an object class from some particular viewpoint or detecting a specific object from an arbitrary view are advancing toward maturity [3, 9, 11], solutions to the problem of object class detection using multiple views are still relatively far behind. Object detection can be considered even more difficult than classification, since it is expected to provide accurate location and size of the object. Researchers in computer vision have studied the problem of multi-view object class detection resulting successful approaches following two major directions. One path attempts to use increasing number of local features by applying multiple feature detectors simultaneously [1, 6, 13, 14, 15]. It has been shown that the recognition performance can be benefited by providing more feature support. However, the spatial connections of the features in each view and/or between different views have not been pursued in these works. These connections can be crucial in object class detection tasks. Recently, much attention has been drawn to the second direction related to multiple views for object class detection [5, 7, 8]. The early methods apply several singleview detectors independently and combine their responses via some arbitration logic. Features are shared among the different single-view detectors to limit the computational overload. Most recently, Thomas et al. [16] developed a single integrated multi-view detector that accumulates evidence from different training views. Their work combines a multi-view specific object recognition system [9], and the Implicit Shape Model for object class detection [11], where single-view codebooks are strongly connected by the exchange of information via sophisticated activation links between each other. In this paper, we present a unified method to relate multiple 2D views based on 3D object modeling. The main contribution of our work is an efficient object detection system capable of recognizing and localizing objects from the same class under different viewing conditions. Thus, 3D locations of the features are considered during detection and better accuracy is obtained. 3. Construction of 3D Shape Model A recently proposed homographic framework was employed in our work to construct 3D models from multiple 2D views [10]. A summary of the 3D reconstruction algorithmisprovidedasfollows View Warping and Intersection Let I i denote the foreground likelihood map (where each pixel value is the likelihood of that pixel being a foreground) in the ith view of total M views. Considering a reference plane, π r, in the scene with homography H πr,i from the ith view to π r,warpingi i to π r gives the warped foreground likelihood map: Î i,r =[H πr,i]i i. (1) The visual hull intersection on π r (AND-fusion of the shadow regions) is achieved by multiplying these warped foreground likelihood maps: θ r = MY Î i,r, (2) i=1 where θ r is the grid of the object occupancy likelihoods plane π r. Each value in θ r gives the likelihood of this grid location being inside the body of the object, indeed, representing a slice of the object cut out by plane π r. It should be noted that due to the multiplication step in (2), the locations outside the visual hull intersection region will be penalized, thus, having a much lower occupancy likelihood Construction in Parallel Planes The grid of the object occupancy likelihood can be computed at an arbitrary number of planes in the scene with different heights, each giving a slice of the object. Naturally this does not apply to planes that do not pass through the object s body, since visual hull intersection on these planes will be empty, therefore a separate check is not necessary. Let v x, v y,andv z denote the vanishing points for the X, Y, and Z directions, respectively, and l be the normalized vanishing line of reference plane in the XY Z coordinate space. The reference plane to the image view homography can be represented as Ĥ ref =[v x v y l]. (3) Supposing that another plane π has a translation of z along the reference direction Z from the reference plane, it is easy

3 (a) Volume projection (b) Plane transformation Figure 1. Illustration of equivalence of 3D to 2D projection and plane transformation using homographies. (a) A 2D view of a 3D volume V is generated by projecting the volume on a image plane. (b) The same view can be obtained by integrating the transformation of each slice in the volume to the image plane using homographies. to show that the homography of plane π to the image view can be computed by Ĥ π =[v x v y αzv z + l] =Ĥ ref +[I αzv z ], (4) where α is a scaling factor. The image to plane homography H π is obtained by inverting Ĥ π. Starting with a reference plane in the scene (typically the ground plane), visual hull intersection is performed on successively parallel planes in the up direction along the body of the object. The occupancy grids θ i are stacked up to create a three dimensional data structure Θ =[θ 1 ; θ 2 ;...θ M ]. Θ represents a discrete sampling of a continuous occupancy space encapsulating the object shape. Object structure is then segmented out from Θ by dividing the space into the object and background regions using the geodesic active contour method [2]. By using the above homography based framework, 3D models for different objects can be constructed. 4. 3D Feature Model Description and Training In the proposed method, not only the 3D shape of the target object is exploited, but also the appearance features. We relate the features with the 3D model to construct a feature model for object class detection Attaching 2D Features to 3D Model The features used in our work are computed using the SIFT feature detector [12]. Feature vectors are computed for all of the training images. In order to efficiently relate the features computed from different views and different objects, all the detected features are attached to the 3D surface of the previously built model. By using the 3D feature model, we avoid storing all the 2D training views, thus there is no need to build complicated connections between the views. The spatial relationship between the feature points from different views are readily available, which can be easily retrieved when matched feature points are found. The features computed in 2D images are attached to the 3D model by using the novel homographic framework. Instead of directly finding the 3D location of each 2D feature, we map the 3D points from the model s surface to the 2D views, and find the corresponding features. Our method does not require the estimation of a projection matrix from 3D model to a 2D image plane, which is a non-trivial problem. In our work, the problem is successfully solved by transforming the model to various image planes using homography. Since the homographies between the model and the image planes have already been obtained during the construction of the 3D model, we are able to map the 3D points to 2D planes using homography transformation. In our work, a 3D shape is represented by a binary volume V, which consists of K slices S j, j [1,K]. As shown in Fig. 1(b), each slice of the object is transformed to a 2D image plane by using the corresponding homography Ĥ in (4). The transformed slice accounts for a small patch of the object projection. Integrating all these K patches together, the whole projection of 3D object in the 2D image plane can be produced. In this way, we obtain the model projection by using a series of simple homography transformations and the hard problem of estimating the projection matrix of a 3D model to a 2D view is avoided. In our method, the 3D shapes are represented using binary volumes with a stack of slices along the reference direction. Thus, the surface points can be easily obtained by applying edge detection techniques. After transforming the surface points to 2D planes, feature vectors computed in 2D can be related to the 3D points according to their locations. That is the way a 3D feature model is built Beyond the Model Views The training images in our work come from two sources. One set of images is taken around a specific object of the target class to reconstruct it in 3D as shown in Fig. 2. These images are called model views, which provide multiple views of the object but are limited to the specific object. To generalize the model for recognizing other objects in the same class, another set of training images is obtained by using Google image search. Images of objects in the same class with different appearances and postures are selected. These images are denoted as the supplemental views. Since the homographies between the supplemental images and the 3D model are unknown, features computed from the supplemental images cannot be directly attached to the feature model. Instead, we utilize the model views as bridges to connect the supplemental images to the model as illustrated in Fig. 2. For each supplemental image, the model view, which has the most similar viewpoint is specified. The supplemental images are deformed to their specified view by using an affine transformation alignment. Then we can assume that each supplemental image will have the

4 Figure 2. Construction of 3D feature model for motorbikes. 3D shape model of motorbike (at center) is constructed using the model views (images on the inner circle) taken around the object from different viewpoints. Supplemental images (outer circle) of different motorbikes are obtained by using Google s image search. The supplemental images are aligned with the model views for feature mapping. Feature vectors are computed from all the training images and then attached to the 3D model surface by using the homography transformation. same homography as the model s corresponding view. The 2D features computed from all of the supplemental training images can now be correctly attached to the 3D model surface using the same method as discussed for the model views. A codebook is constructed by combining all the mapped features with their 3D locations. 5. Object Class Detection Given a new test image, our objective is to detect objects belonging to the same class in this image by using the learnt 3D feature model M. Each entry of M consists of a code and its 3D locations {c, l 3 c}. LetF denote the SIFT features computed from the input image, which is composed by the feature descriptor and its 2D location in the image {f,l 2 f }. Object O n is detected by matching the features F to the 3D feature model M. In our work, feature matching is achieved in three phases. In the first phase, we match the features by comparing all the input features to the codebook entries in Euclidean space. However, not all the matched codebook entries in 3D are visible at the same time from a particular viewpoint. So, in the second phase, matched codes in 3D are projected to viewing planes and hypotheses of viewpoints are made by selecting viewing planes with the largest number of visible points projected. In the third phase, for each hypothesis, the projected points are compared to 2D matched feature points using both feature descriptors and locations. This is done by iteratively estimating the affine transformation between the feature point sets and removing the outliers with large distance between corresponding points. Outliers belonging to the background can be rejected during this matching process. The object location and bounding box is then determined according to the 2D locations of the final matched feature points. The confidence of detection is given by the degree of match. 6. Experimental Results The proposed method has been tested on two object classes: motorbikes and horses. For the motorbikes, we took 23 model views around a motorbike and obtained 45

5 Figure 3. Detection of motorbikes and horses using the proposed approach. The ground truth is shown in green and red boxes display our detected results. supplemental views by using Google s image search. Some training images of the motorbikes and the 3D shape model are shown in Fig. 2. For the horses, 18 model views were taken and 51 supplemental views were obtained. To measure the performance of our 3D feature model based object class detection technique, we have evaluated the method on the PASCAL VOC Challenge 2006 test dataset [4], which has become a standard testing dataset for objective evaluation of object classification and detection algorithms. The dataset is very challenging due to the large variability in the scale and poses, the extensive clutter, and poor imaging conditions. Some successful detection results are shown in Fig. 3. The green box indicates the ground truth, while our results are shown in red boxes. For quantitative evaluation, we adopt the same evaluation criteria used in PASCAL VOC challenge, so that our results can be directly comparable with [8, 16, 4]. By using this criteria, a detection is considered correct, if the area of overlap between the predicted bounding box B p and ground truth bounding box B gt exceeds 50% using the formula area(b p B gt ) > 0.5. (5) area(b p B gt ) The average precision (AP) and precision-recall (PR) curve can then be computed for performance evaluation. Fig. 4(a) shows the PR curves of our approach and the methods in [8, 16] for motorbike detection. The curve of our approach shows a substantial improvement over the precision compared to the method in [8], which is also indicated by the AP value (0.182). Although our performance is lower than that of [16], considering the smaller training image set used in our experiments, this can be regarded as satisfactory. Fig. 4(b) shows the performance curves for horse detection. While there is no result reported in the VOC challenge using researchers own training dataset for this task, we compared our result to those using the provided training dataset. Our approach performs better than the reported methods and obtained AP value of It is noted that the absolute performance level is lower than that of motorbike detection, which might be caused by the non-rigid body deformation of horses. 7. Conclusion In this paper, we have proposed a novel multi-view generic object class detection method based on 3D object

6 Figure 4. The PR curves for (a) motorbike detection and (b) horse detection using our 3D feature model based approach. The curves reported in [4] on the same test dataset are also included for comparison. shape and appearance modeling. The proposed method builds a 3D feature model for establishing spatial connections between different image views by mapping appearance features to the surfaces of a 3D shape. Through a novel homographic framework, our method exploits the relationship between multiple 2D views in a more unified manner. Experimental evaluation of the proposed method suggests collaborative information in the 2D training images can be represented in a more unified way through a 3D feature model of the object. We have also revealed that both appearance and shape can be salient properties to assist in object detection. Performance of the proposed method has been evaluated using the PASCAL VOC challenge dataset and promising results have been demonstrated. In our future work, we plan to extend our method by taking supplemental views in a more automated fashion. So, more supplemental views can be easily incorporated to improve the performance. We will also test the method on objects classes with more complex shapes and appearances. Acknowledgments This research was funded in part by the U.S. Government VACE program. References [1] A. C. Berg, T. L. Berg, and J. Malik. Shape matching and object recognition using low distortion correspondences. In CVPR (1), pages 26 33, [2] V. Caselles, R. Kimmel, and G. Sapiro. Geodesic active contours. International Journal of Computer Vision, 22(1):61 79, [3] H. Chang and D.-Y. Yeung. Graph laplacian kernels for object classification from a single example. In CVPR (2), pages , [4] M. Everingham, A. Zisserman, C. Williams, and L. van Gool. The PASCAL Visual Object Classes Challenge 2006 (VOC2006) Results. 5, 6 [5] J. D. R. Farquhar, D. R. Hardoon, H. Meng, J. Shawe-Taylor, and S. Szedmák. Two view learning: SVM-2K, theory and practice. In NIPS, [6] L. Fei-Fei, R. Fergus, and P. Perona. A bayesian approach to unsupervised one-shot learning of object categories. In ICCV, pages , [7] R. Fergus, L. Fei-Fei, P. Perona, and A. Zisserman. Learning object categories from google s image search. In ICCV, pages , [8] R. Fergus, P. Perona, and A. Zisserman. A sparse object category model for efficient learning and exhaustive recognition. In CVPR, pages , , 5 [9] V. Ferrari, T. Tuytelaars, and L. Van Gool. Integrating multiple model views for object recognition. In CVPR,volume2, pages , [10] S. M. Khan, P. Yan, and M. Shah. A homographic framework for the fusion of multi-view silhouettes. In ICCV, , 2 [11] B. Leibe and B. Schiele. Scale-invariant object categorization using a scale-adaptive mean-shift search. In DAGM, pages , [12] D. G. Lowe. Distinctive image features from scale-invariant keypoints. International Journal of Computer Vision, 60:91 110, Nov , 3 [13] E. Nowak, F. Jurie, and B. Triggs. Sampling strategies for bag-of-features image classification. In ECCV (4), pages , [14] J. Shotton, A. Blake, and R. Cipolla. Contour-based learning for object detection. In ICCV, pages , [15] E. B. Sudderth, A. B. Torralba, W. T. Freeman, and A. S. Willsky. Learning hierarchical models of scenes, objects, and parts. In ICCV, pages , [16] A. Thomas, V. Ferrari, B. Leibe, T. Tuytelaars, B. Schiele, and L. Van Gool. Towards multi-view object class detection. In CVPR, volume 2, pages , , 5

Build Panoramas on Android Phones

Build Panoramas on Android Phones Build Panoramas on Android Phones Tao Chu, Bowen Meng, Zixuan Wang Stanford University, Stanford CA Abstract The purpose of this work is to implement panorama stitching from a sequence of photos taken

More information

Object class recognition using unsupervised scale-invariant learning

Object class recognition using unsupervised scale-invariant learning Object class recognition using unsupervised scale-invariant learning Rob Fergus Pietro Perona Andrew Zisserman Oxford University California Institute of Technology Goal Recognition of object categories

More information

Multi-View Object Class Detection with a 3D Geometric Model

Multi-View Object Class Detection with a 3D Geometric Model Multi-View Object Class Detection with a 3D Geometric Model Joerg Liebelt IW-SI, EADS Innovation Works D-81663 Munich, Germany joerg.liebelt@eads.net Cordelia Schmid LEAR, INRIA Grenoble F-38330 Montbonnot,

More information

Object Recognition. Selim Aksoy. Bilkent University saksoy@cs.bilkent.edu.tr

Object Recognition. Selim Aksoy. Bilkent University saksoy@cs.bilkent.edu.tr Image Classification and Object Recognition Selim Aksoy Department of Computer Engineering Bilkent University saksoy@cs.bilkent.edu.tr Image classification Image (scene) classification is a fundamental

More information

Learning Motion Categories using both Semantic and Structural Information

Learning Motion Categories using both Semantic and Structural Information Learning Motion Categories using both Semantic and Structural Information Shu-Fai Wong, Tae-Kyun Kim and Roberto Cipolla Department of Engineering, University of Cambridge, Cambridge, CB2 1PZ, UK {sfw26,

More information

Local features and matching. Image classification & object localization

Local features and matching. Image classification & object localization Overview Instance level search Local features and matching Efficient visual recognition Image classification & object localization Category recognition Image classification: assigning a class label to

More information

The goal is multiply object tracking by detection with application on pedestrians.

The goal is multiply object tracking by detection with application on pedestrians. Coupled Detection and Trajectory Estimation for Multi-Object Tracking By B. Leibe, K. Schindler, L. Van Gool Presented By: Hanukaev Dmitri Lecturer: Prof. Daphna Wienshall The Goal The goal is multiply

More information

Probabilistic Latent Semantic Analysis (plsa)

Probabilistic Latent Semantic Analysis (plsa) Probabilistic Latent Semantic Analysis (plsa) SS 2008 Bayesian Networks Multimedia Computing, Universität Augsburg Rainer.Lienhart@informatik.uni-augsburg.de www.multimedia-computing.{de,org} References

More information

VEHICLE LOCALISATION AND CLASSIFICATION IN URBAN CCTV STREAMS

VEHICLE LOCALISATION AND CLASSIFICATION IN URBAN CCTV STREAMS VEHICLE LOCALISATION AND CLASSIFICATION IN URBAN CCTV STREAMS Norbert Buch 1, Mark Cracknell 2, James Orwell 1 and Sergio A. Velastin 1 1. Kingston University, Penrhyn Road, Kingston upon Thames, KT1 2EE,

More information

Object Categorization using Co-Occurrence, Location and Appearance

Object Categorization using Co-Occurrence, Location and Appearance Object Categorization using Co-Occurrence, Location and Appearance Carolina Galleguillos Andrew Rabinovich Serge Belongie Department of Computer Science and Engineering University of California, San Diego

More information

Using geometry and related things

Using geometry and related things Using geometry and related things Region labels + Boundaries and objects Stronger geometric constraints from domain knowledge Reasoning on aspects and poses 3D point clouds Qualitative More quantitative

More information

The use of computer vision technologies to augment human monitoring of secure computing facilities

The use of computer vision technologies to augment human monitoring of secure computing facilities The use of computer vision technologies to augment human monitoring of secure computing facilities Marius Potgieter School of Information and Communication Technology Nelson Mandela Metropolitan University

More information

Recognizing Cats and Dogs with Shape and Appearance based Models. Group Member: Chu Wang, Landu Jiang

Recognizing Cats and Dogs with Shape and Appearance based Models. Group Member: Chu Wang, Landu Jiang Recognizing Cats and Dogs with Shape and Appearance based Models Group Member: Chu Wang, Landu Jiang Abstract Recognizing cats and dogs from images is a challenging competition raised by Kaggle platform

More information

Efficient visual search of local features. Cordelia Schmid

Efficient visual search of local features. Cordelia Schmid Efficient visual search of local features Cordelia Schmid Visual search change in viewing angle Matches 22 correct matches Image search system for large datasets Large image dataset (one million images

More information

Behavior Analysis in Crowded Environments. XiaogangWang Department of Electronic Engineering The Chinese University of Hong Kong June 25, 2011

Behavior Analysis in Crowded Environments. XiaogangWang Department of Electronic Engineering The Chinese University of Hong Kong June 25, 2011 Behavior Analysis in Crowded Environments XiaogangWang Department of Electronic Engineering The Chinese University of Hong Kong June 25, 2011 Behavior Analysis in Sparse Scenes Zelnik-Manor & Irani CVPR

More information

Online Learning of Patch Perspective Rectification for Efficient Object Detection

Online Learning of Patch Perspective Rectification for Efficient Object Detection Online Learning of Patch Perspective Rectification for Efficient Object Detection Stefan Hinterstoisser 1, Selim Benhimane 1, Nassir Navab 1, Pascal Fua 2, Vincent Lepetit 2 1 Department of Computer Science,

More information

Scale-Invariant Object Categorization using a Scale-Adaptive Mean-Shift Search

Scale-Invariant Object Categorization using a Scale-Adaptive Mean-Shift Search in DAGM 04 Pattern Recognition Symposium, Tübingen, Germany, Aug 2004. Scale-Invariant Object Categorization using a Scale-Adaptive Mean-Shift Search Bastian Leibe ½ and Bernt Schiele ½¾ ½ Perceptual Computing

More information

Fast Matching of Binary Features

Fast Matching of Binary Features Fast Matching of Binary Features Marius Muja and David G. Lowe Laboratory for Computational Intelligence University of British Columbia, Vancouver, Canada {mariusm,lowe}@cs.ubc.ca Abstract There has been

More information

Vehicle Tracking by Simultaneous Detection and Viewpoint Estimation

Vehicle Tracking by Simultaneous Detection and Viewpoint Estimation Vehicle Tracking by Simultaneous Detection and Viewpoint Estimation Ricardo Guerrero-Gómez-Olmedo, Roberto López-Sastre, Saturnino Maldonado-Bascón, and Antonio Fernández-Caballero 2 GRAM, Department of

More information

Automatic 3D Reconstruction via Object Detection and 3D Transformable Model Matching CS 269 Class Project Report

Automatic 3D Reconstruction via Object Detection and 3D Transformable Model Matching CS 269 Class Project Report Automatic 3D Reconstruction via Object Detection and 3D Transformable Model Matching CS 69 Class Project Report Junhua Mao and Lunbo Xu University of California, Los Angeles mjhustc@ucla.edu and lunbo

More information

WEIGHTED BAG OF VISUAL WORDS FOR OBJECT RECOGNITION

WEIGHTED BAG OF VISUAL WORDS FOR OBJECT RECOGNITION WEIGHTED BAG OF VISUAL WORDS FOR OBJECT RECOGNITION Marco San Biagio1*, Loris Bazzani1,2*, Marco Cristani1,2, Vittorio Murino1,2 1 Pattern Analysis & Computer Vision, Istituto Italiano di Tecnologia, Via

More information

Comparison of Non-linear Dimensionality Reduction Techniques for Classification with Gene Expression Microarray Data

Comparison of Non-linear Dimensionality Reduction Techniques for Classification with Gene Expression Microarray Data CMPE 59H Comparison of Non-linear Dimensionality Reduction Techniques for Classification with Gene Expression Microarray Data Term Project Report Fatma Güney, Kübra Kalkan 1/15/2013 Keywords: Non-linear

More information

Camera geometry and image alignment

Camera geometry and image alignment Computer Vision and Machine Learning Winter School ENS Lyon 2010 Camera geometry and image alignment Josef Sivic http://www.di.ens.fr/~josef INRIA, WILLOW, ENS/INRIA/CNRS UMR 8548 Laboratoire d Informatique,

More information

Computer Vision - part II

Computer Vision - part II Computer Vision - part II Review of main parts of Section B of the course School of Computer Science & Statistics Trinity College Dublin Dublin 2 Ireland www.scss.tcd.ie Lecture Name Course Name 1 1 2

More information

A Comparative Study between SIFT- Particle and SURF-Particle Video Tracking Algorithms

A Comparative Study between SIFT- Particle and SURF-Particle Video Tracking Algorithms A Comparative Study between SIFT- Particle and SURF-Particle Video Tracking Algorithms H. Kandil and A. Atwan Information Technology Department, Faculty of Computer and Information Sciences, Mansoura University,El-Gomhoria

More information

Jiří Matas. Hough Transform

Jiří Matas. Hough Transform Hough Transform Jiří Matas Center for Machine Perception Department of Cybernetics, Faculty of Electrical Engineering Czech Technical University, Prague Many slides thanks to Kristen Grauman and Bastian

More information

CS 534: Computer Vision 3D Model-based recognition

CS 534: Computer Vision 3D Model-based recognition CS 534: Computer Vision 3D Model-based recognition Ahmed Elgammal Dept of Computer Science CS 534 3D Model-based Vision - 1 High Level Vision Object Recognition: What it means? Two main recognition tasks:!

More information

More Local Structure Information for Make-Model Recognition

More Local Structure Information for Make-Model Recognition More Local Structure Information for Make-Model Recognition David Anthony Torres Dept. of Computer Science The University of California at San Diego La Jolla, CA 9093 Abstract An object classification

More information

Simultaneous Gamma Correction and Registration in the Frequency Domain

Simultaneous Gamma Correction and Registration in the Frequency Domain Simultaneous Gamma Correction and Registration in the Frequency Domain Alexander Wong a28wong@uwaterloo.ca William Bishop wdbishop@uwaterloo.ca Department of Electrical and Computer Engineering University

More information

Randomized Trees for Real-Time Keypoint Recognition

Randomized Trees for Real-Time Keypoint Recognition Randomized Trees for Real-Time Keypoint Recognition Vincent Lepetit Pascal Lagger Pascal Fua Computer Vision Laboratory École Polytechnique Fédérale de Lausanne (EPFL) 1015 Lausanne, Switzerland Email:

More information

Object Detection in Multi-View X-Ray Images

Object Detection in Multi-View X-Ray Images Object Detection in Multi-View X-Ray Images Thorsten Franzel, Uwe Schmidt, and Stefan Roth Department of Computer Science, TU Darmstadt Abstract. Motivated by aiding human operators in the detection of

More information

Modified Sift Algorithm for Appearance Based Recognition of American Sign Language

Modified Sift Algorithm for Appearance Based Recognition of American Sign Language Modified Sift Algorithm for Appearance Based Recognition of American Sign Language Jaspreet Kaur,Navjot Kaur Electronics and Communication Engineering Department I.E.T. Bhaddal, Ropar, Punjab,India. Abstract:

More information

Edge-based Template Matching and Tracking for Perspectively Distorted Planar Objects

Edge-based Template Matching and Tracking for Perspectively Distorted Planar Objects Edge-based Template Matching and Tracking for Perspectively Distorted Planar Objects Andreas Hofhauser, Carsten Steger, and Nassir Navab TU München, Boltzmannstr. 3, 85748 Garching bei München, Germany

More information

Robust Panoramic Image Stitching

Robust Panoramic Image Stitching Robust Panoramic Image Stitching CS231A Final Report Harrison Chau Department of Aeronautics and Astronautics Stanford University Stanford, CA, USA hwchau@stanford.edu Robert Karol Department of Aeronautics

More information

Group Sparse Coding. Fernando Pereira Google Mountain View, CA pereira@google.com. Dennis Strelow Google Mountain View, CA strelow@google.

Group Sparse Coding. Fernando Pereira Google Mountain View, CA pereira@google.com. Dennis Strelow Google Mountain View, CA strelow@google. Group Sparse Coding Samy Bengio Google Mountain View, CA bengio@google.com Fernando Pereira Google Mountain View, CA pereira@google.com Yoram Singer Google Mountain View, CA singer@google.com Dennis Strelow

More information

Learning to Segment. Eran Borenstein and Shimon Ullman

Learning to Segment. Eran Borenstein and Shimon Ullman Learning to Segment Eran Borenstein and Shimon Ullman Faculty of Mathematics and Computer Science Weizmann Institute of Science Rehovot, Israel 76100 {eran.borenstein,shimon.ullman}@weizmann.ac.il Abstract.

More information

IMPLICIT SHAPE MODELS FOR OBJECT DETECTION IN 3D POINT CLOUDS

IMPLICIT SHAPE MODELS FOR OBJECT DETECTION IN 3D POINT CLOUDS IMPLICIT SHAPE MODELS FOR OBJECT DETECTION IN 3D POINT CLOUDS Alexander Velizhev 1 (presenter) Roman Shapovalov 2 Konrad Schindler 3 1 Hexagon Technology Center, Heerbrugg, Switzerland 2 Graphics & Media

More information

Are we ready for Autonomous Driving? The KITTI Vision Benchmark Suite

Are we ready for Autonomous Driving? The KITTI Vision Benchmark Suite Are we ready for Autonomous Driving? The KITTI Vision Benchmark Suite Philip Lenz 1 Andreas Geiger 2 Christoph Stiller 1 Raquel Urtasun 3 1 KARLSRUHE INSTITUTE OF TECHNOLOGY 2 MAX-PLANCK-INSTITUTE IS 3

More information

Recognition Using Visual Phrases

Recognition Using Visual Phrases Recognition Using Visual Phrases Mohammad Amin Sadeghi 1,2, Ali Farhadi 1 1 Computer Science Department, University of Illinois at Urbana-Champaign 2 Computer Vision Group, Institute for Research in Fundamental

More information

Pedestrian Detection with RCNN

Pedestrian Detection with RCNN Pedestrian Detection with RCNN Matthew Chen Department of Computer Science Stanford University mcc17@stanford.edu Abstract In this paper we evaluate the effectiveness of using a Region-based Convolutional

More information

Polygonal Approximation of Closed Curves across Multiple Views

Polygonal Approximation of Closed Curves across Multiple Views Polygonal Approximation of Closed Curves across Multiple Views M. Pawan Kumar Saurabh Goyal C. V. Jawahar P. J. Narayanan Centre for Visual Information Technology International Institute of Information

More information

Visual Data Combination for Object Detection and Localization for Autonomous Robot Manipulation Tasks

Visual Data Combination for Object Detection and Localization for Autonomous Robot Manipulation Tasks Visual Data Combination for Object Detection and Localization for Autonomous Robot Manipulation Tasks Luis A. Morgado-Ramirez, Sergio Hernandez-Mendez, Luis F. Marin- Urias, Antonio Marin-Hernandez, Homero

More information

TouchPaper - An Augmented Reality Application with Cloud-Based Image Recognition Service

TouchPaper - An Augmented Reality Application with Cloud-Based Image Recognition Service TouchPaper - An Augmented Reality Application with Cloud-Based Image Recognition Service Feng Tang, Daniel R. Tretter, Qian Lin HP Laboratories HPL-2012-131R1 Keyword(s): image recognition; cloud service;

More information

Fast field survey with a smartphone

Fast field survey with a smartphone Fast field survey with a smartphone A. Masiero F. Fissore, F. Pirotti, A. Guarnieri, A. Vettore CIRGEO Interdept. Research Center of Geomatics University of Padova Italy cirgeo@unipd.it 1 Mobile Mapping

More information

Segmentation as Selective Search for Object Recognition

Segmentation as Selective Search for Object Recognition Segmentation as Selective Search for Object Recognition Koen E. A. van de Sande Jasper R. R. Uijlings Theo Gevers Arnold W. M. Smeulders University of Amsterdam University of Trento Amsterdam, The Netherlands

More information

Epipolar Geometry Prof. D. Stricker

Epipolar Geometry Prof. D. Stricker Outline 1. Short introduction: points and lines Epipolar Geometry Prof. D. Stricker 2. Two views geometry: Epipolar geometry Relation point/line in two views The geometry of two cameras Definition of the

More information

A Study on SURF Algorithm and Real-Time Tracking Objects Using Optical Flow

A Study on SURF Algorithm and Real-Time Tracking Objects Using Optical Flow , pp.233-237 http://dx.doi.org/10.14257/astl.2014.51.53 A Study on SURF Algorithm and Real-Time Tracking Objects Using Optical Flow Giwoo Kim 1, Hye-Youn Lim 1 and Dae-Seong Kang 1, 1 Department of electronices

More information

Learning 3D Object Templates by Hierarchical Quantization of Geometry and Appearance Spaces

Learning 3D Object Templates by Hierarchical Quantization of Geometry and Appearance Spaces Learning 3D Object Templates by Hierarchical Quantization of Geometry and Appearance Spaces Wenze Hu Department of Statistics, UCLA wzhu@stat.ucla.edu Abstract This paper presents a method for learning

More information

Wii Remote Calibration Using the Sensor Bar

Wii Remote Calibration Using the Sensor Bar Wii Remote Calibration Using the Sensor Bar Alparslan Yildiz Abdullah Akay Yusuf Sinan Akgul GIT Vision Lab - http://vision.gyte.edu.tr Gebze Institute of Technology Kocaeli, Turkey {yildiz, akay, akgul}@bilmuh.gyte.edu.tr

More information

Discovering objects and their location in images

Discovering objects and their location in images Discovering objects and their location in images Josef Sivic Bryan C. Russell Alexei A. Efros Andrew Zisserman William T. Freeman Dept. of Engineering Science CS and AI Laboratory School of Computer Science

More information

High Level Describable Attributes for Predicting Aesthetics and Interestingness

High Level Describable Attributes for Predicting Aesthetics and Interestingness High Level Describable Attributes for Predicting Aesthetics and Interestingness Sagnik Dhar Vicente Ordonez Tamara L Berg Stony Brook University Stony Brook, NY 11794, USA tlberg@cs.stonybrook.edu Abstract

More information

An Image-Based System for Urban Navigation

An Image-Based System for Urban Navigation An Image-Based System for Urban Navigation Duncan Robertson and Roberto Cipolla Cambridge University Engineering Department Trumpington Street, Cambridge, CB2 1PZ, UK Abstract We describe the prototype

More information

Lecture 2: Homogeneous Coordinates, Lines and Conics

Lecture 2: Homogeneous Coordinates, Lines and Conics Lecture 2: Homogeneous Coordinates, Lines and Conics 1 Homogeneous Coordinates In Lecture 1 we derived the camera equations λx = P X, (1) where x = (x 1, x 2, 1), X = (X 1, X 2, X 3, 1) and P is a 3 4

More information

EFFICIENT VEHICLE TRACKING AND CLASSIFICATION FOR AN AUTOMATED TRAFFIC SURVEILLANCE SYSTEM

EFFICIENT VEHICLE TRACKING AND CLASSIFICATION FOR AN AUTOMATED TRAFFIC SURVEILLANCE SYSTEM EFFICIENT VEHICLE TRACKING AND CLASSIFICATION FOR AN AUTOMATED TRAFFIC SURVEILLANCE SYSTEM Amol Ambardekar, Mircea Nicolescu, and George Bebis Department of Computer Science and Engineering University

More information

Mean-Shift Tracking with Random Sampling

Mean-Shift Tracking with Random Sampling 1 Mean-Shift Tracking with Random Sampling Alex Po Leung, Shaogang Gong Department of Computer Science Queen Mary, University of London, London, E1 4NS Abstract In this work, boosting the efficiency of

More information

The Visual Internet of Things System Based on Depth Camera

The Visual Internet of Things System Based on Depth Camera The Visual Internet of Things System Based on Depth Camera Xucong Zhang 1, Xiaoyun Wang and Yingmin Jia Abstract The Visual Internet of Things is an important part of information technology. It is proposed

More information

Augmented Reality Tic-Tac-Toe

Augmented Reality Tic-Tac-Toe Augmented Reality Tic-Tac-Toe Joe Maguire, David Saltzman Department of Electrical Engineering jmaguire@stanford.edu, dsaltz@stanford.edu Abstract: This project implements an augmented reality version

More information

Segmentation. Lecture 12. Many slides from: S. Lazebnik, K. Grauman and P. Kumar

Segmentation. Lecture 12. Many slides from: S. Lazebnik, K. Grauman and P. Kumar Segmentation Lecture 12 Many slides from: S. Lazebnik, K. Grauman and P. Kumar Image Segmentation Image segmentation The goals of segmentation Group together similar-looking pixels for efficiency of further

More information

Pictorial Structures Revisited: People Detection and Articulated Pose Estimation

Pictorial Structures Revisited: People Detection and Articulated Pose Estimation Pictorial Structures Revisited: People Detection and Articulated Pose Estimation Mykhaylo Andriluka, Stefan Roth, and Bernt Schiele Department of Computer Science, TU Darmstadt Abstract Non-rigid object

More information

Scalable Object Detection by Filter Compression with Regularized Sparse Coding

Scalable Object Detection by Filter Compression with Regularized Sparse Coding Scalable Object Detection by Filter Compression with Regularized Sparse Coding Ting-Hsuan Chao, Yen-Liang Lin, Yin-Hsi Kuo, and Winston H Hsu National Taiwan University, Taipei, Taiwan Abstract For practical

More information

Distinctive Image Features from Scale-Invariant Keypoints

Distinctive Image Features from Scale-Invariant Keypoints Distinctive Image Features from Scale-Invariant Keypoints David G. Lowe Computer Science Department University of British Columbia Vancouver, B.C., Canada lowe@cs.ubc.ca January 5, 2004 Abstract This paper

More information

Automatic Labeling of Lane Markings for Autonomous Vehicles

Automatic Labeling of Lane Markings for Autonomous Vehicles Automatic Labeling of Lane Markings for Autonomous Vehicles Jeffrey Kiske Stanford University 450 Serra Mall, Stanford, CA 94305 jkiske@stanford.edu 1. Introduction As autonomous vehicles become more popular,

More information

Classification using intersection kernel SVMs is efficient

Classification using intersection kernel SVMs is efficient Classification using intersection kernel SVMs is efficient Jitendra Malik UC Berkeley Joint work with Subhransu Maji and Alex Berg Fast intersection kernel SVMs and other generalizations of linear SVMs

More information

Cees Snoek. Machine. Humans. Multimedia Archives. Euvision Technologies The Netherlands. University of Amsterdam The Netherlands. Tree.

Cees Snoek. Machine. Humans. Multimedia Archives. Euvision Technologies The Netherlands. University of Amsterdam The Netherlands. Tree. Visual search: what's next? Cees Snoek University of Amsterdam The Netherlands Euvision Technologies The Netherlands Problem statement US flag Tree Aircraft Humans Dog Smoking Building Basketball Table

More information

Multiview Social Behavior Analysis in Work Environments

Multiview Social Behavior Analysis in Work Environments Multiview Social Behavior Analysis in Work Environments Chih-Wei Chen Department of Electrical Engineering Stanford University Stanford, California 94305 Email: louistw@stanford.edu Hamid Aghajan Department

More information

Shape Measurement of a Sewer Pipe. Using a Mobile Robot with Computer Vision

Shape Measurement of a Sewer Pipe. Using a Mobile Robot with Computer Vision International Journal of Advanced Robotic Systems ARTICLE Shape Measurement of a Sewer Pipe Using a Mobile Robot with Computer Vision Regular Paper Kikuhito Kawasue 1,* and Takayuki Komatsu 1 1 Department

More information

GPS-aided Recognition-based User Tracking System with Augmented Reality in Extreme Large-scale Areas

GPS-aided Recognition-based User Tracking System with Augmented Reality in Extreme Large-scale Areas GPS-aided Recognition-based User Tracking System with Augmented Reality in Extreme Large-scale Areas Wei Guan Computer Graphics and Immersive Technologies Computer Science, USC wguan@usc.edu Suya You Computer

More information

CATEGORIZATION OF SIMILAR OBJECTS USING BAG OF VISUAL WORDS AND k NEAREST NEIGHBOUR CLASSIFIER

CATEGORIZATION OF SIMILAR OBJECTS USING BAG OF VISUAL WORDS AND k NEAREST NEIGHBOUR CLASSIFIER TECHNICAL SCIENCES Abbrev.: Techn. Sc., No 15(2), Y 2012 CATEGORIZATION OF SIMILAR OBJECTS USING BAG OF VISUAL WORDS AND k NEAREST NEIGHBOUR CLASSIFIER Piotr Artiemjew, Przemysław Górecki, Krzysztof Sopyła

More information

UNSUPERVISED COSEGMENTATION BASED ON SUPERPIXEL MATCHING AND FASTGRABCUT. Hongkai Yu and Xiaojun Qi

UNSUPERVISED COSEGMENTATION BASED ON SUPERPIXEL MATCHING AND FASTGRABCUT. Hongkai Yu and Xiaojun Qi UNSUPERVISED COSEGMENTATION BASED ON SUPERPIXEL MATCHING AND FASTGRABCUT Hongkai Yu and Xiaojun Qi Department of Computer Science, Utah State University, Logan, UT 84322-4205 hongkai.yu@aggiemail.usu.edu

More information

A Learning Based Method for Super-Resolution of Low Resolution Images

A Learning Based Method for Super-Resolution of Low Resolution Images A Learning Based Method for Super-Resolution of Low Resolution Images Emre Ugur June 1, 2004 emre.ugur@ceng.metu.edu.tr Abstract The main objective of this project is the study of a learning based method

More information

The Design and Implementation of Traffic Accident Identification System Based on Video

The Design and Implementation of Traffic Accident Identification System Based on Video 3rd International Conference on Multimedia Technology(ICMT 2013) The Design and Implementation of Traffic Accident Identification System Based on Video Chenwei Xiang 1, Tuo Wang 2 Abstract: With the rapid

More information

Delving into the whorl of flower segmentation

Delving into the whorl of flower segmentation Delving into the whorl of flower segmentation Maria-Elena Nilsback and Andrew Zisserman Visual Geometry Group Deparment of Engineering Sciene, University of Oxford Abstract We describe an algorithm for

More information

Detection and Classification of Multiple Objects using an RGB-D Sensor and Linear Spatial Pyramid Matching

Detection and Classification of Multiple Objects using an RGB-D Sensor and Linear Spatial Pyramid Matching Electronic Letters on Computer Vision and Image Analysis 12(2):78-87; 2013 Detection and Classification of Multiple Objects using an RGB-D Sensor and Linear Spatial Pyramid Matching Michalis Dimitriou

More information

Image Classification using Random Forests and Ferns

Image Classification using Random Forests and Ferns Image Classification using Random Forests and Ferns Anna Bosch Computer Vision Group University of Girona aboschr@eia.udg.es Andrew Zisserman Dept. of Engineering Science University of Oxford az@robots.ox.ac.uk

More information

Find Me: An Android App for Measurement and Localization

Find Me: An Android App for Measurement and Localization Find Me: An Android App for Measurement and Localization Grace Tsai and Sairam Sundaresan Department of Electrical Engineering and Computer Science University of Michigan, Ann Arbor Email: {gstsai,sairams}@umich.edu

More information

Tracking in flussi video 3D. Ing. Samuele Salti

Tracking in flussi video 3D. Ing. Samuele Salti Seminari XXIII ciclo Tracking in flussi video 3D Ing. Tutors: Prof. Tullio Salmon Cinotti Prof. Luigi Di Stefano The Tracking problem Detection Object model, Track initiation, Track termination, Tracking

More information

BRIEF: Binary Robust Independent Elementary Features

BRIEF: Binary Robust Independent Elementary Features BRIEF: Binary Robust Independent Elementary Features Michael Calonder, Vincent Lepetit, Christoph Strecha, and Pascal Fua CVLab, EPFL, Lausanne, Switzerland e-mail: firstname.lastname@epfl.ch Abstract.

More information

3 Image-Based Photo Hulls. 2 Image-Based Visual Hulls. 3.1 Approach. 3.2 Photo-Consistency. Figure 1. View-dependent geometry.

3 Image-Based Photo Hulls. 2 Image-Based Visual Hulls. 3.1 Approach. 3.2 Photo-Consistency. Figure 1. View-dependent geometry. Image-Based Photo Hulls Greg Slabaugh, Ron Schafer Georgia Institute of Technology Center for Signal and Image Processing Atlanta, GA 30332 {slabaugh, rws}@ece.gatech.edu Mat Hans Hewlett-Packard Laboratories

More information

C4 Computer Vision. 4 Lectures Michaelmas Term Tutorial Sheet Prof A. Zisserman. fundamental matrix, recovering ego-motion, applications.

C4 Computer Vision. 4 Lectures Michaelmas Term Tutorial Sheet Prof A. Zisserman. fundamental matrix, recovering ego-motion, applications. C4 Computer Vision 4 Lectures Michaelmas Term 2004 1 Tutorial Sheet Prof A. Zisserman Overview Lecture 1: Stereo Reconstruction I: epipolar geometry, fundamental matrix. Lecture 2: Stereo Reconstruction

More information

Projection Center Calibration for a Co-located Projector Camera System

Projection Center Calibration for a Co-located Projector Camera System Projection Center Calibration for a Co-located Camera System Toshiyuki Amano Department of Computer and Communication Science Faculty of Systems Engineering, Wakayama University Sakaedani 930, Wakayama,

More information

1. Bag of visual words model: recognizing object categories

1. Bag of visual words model: recognizing object categories 1. Bag of visual words model: recognizing object categories 1 1 1 Problem: Image Classification Given: positive training images containing an object class, and negative training images that don t Classify:

More information

Automatic 3D Object Segmentation in Multiple Views using Volumetric Graph-Cuts

Automatic 3D Object Segmentation in Multiple Views using Volumetric Graph-Cuts Automatic 3D Object Segmentation in Multiple Views using Volumetric Graph-Cuts Neill Campbell 1 George Vogiatzis 2 Carlos Hernández 2 Roberto Cipolla 1 1 Department of Engineering, University of Cambridge,

More information

Introduction to Computer Vision. Week 11, Fall 2010 Instructor: Prof. Ko Nishino

Introduction to Computer Vision. Week 11, Fall 2010 Instructor: Prof. Ko Nishino Introduction to Computer Vision Week 11, Fall 2010 Instructor: Prof. Ko Nishino The Projective Plane Why do we need homogeneous coordinates? represent points at infinity, homographies, perspective projection,

More information

FastKeypointRecognitioninTenLinesofCode

FastKeypointRecognitioninTenLinesofCode FastKeypointRecognitioninTenLinesofCode Mustafa Özuysal Pascal Fua Vincent Lepetit Computer Vision Laboratory École Polytechnique Fédérale de Lausanne(EPFL) 115 Lausanne, Switzerland Email: {Mustafa.Oezuysal,

More information

On Multifont Character Classification in Telugu

On Multifont Character Classification in Telugu On Multifont Character Classification in Telugu Venkat Rasagna, K. J. Jinesh, and C. V. Jawahar International Institute of Information Technology, Hyderabad 500032, INDIA. Abstract. A major requirement

More information

Practical Tour of Visual tracking. David Fleet and Allan Jepson January, 2006

Practical Tour of Visual tracking. David Fleet and Allan Jepson January, 2006 Practical Tour of Visual tracking David Fleet and Allan Jepson January, 2006 Designing a Visual Tracker: What is the state? pose and motion (position, velocity, acceleration, ) shape (size, deformation,

More information

Character Image Patterns as Big Data

Character Image Patterns as Big Data 22 International Conference on Frontiers in Handwriting Recognition Character Image Patterns as Big Data Seiichi Uchida, Ryosuke Ishida, Akira Yoshida, Wenjie Cai, Yaokai Feng Kyushu University, Fukuoka,

More information

Visual Categorization with Bags of Keypoints

Visual Categorization with Bags of Keypoints Visual Categorization with Bags of Keypoints Gabriella Csurka, Christopher R. Dance, Lixin Fan, Jutta Willamowski, Cédric Bray Xerox Research Centre Europe 6, chemin de Maupertuis 38240 Meylan, France

More information

Recognition. Sanja Fidler CSC420: Intro to Image Understanding 1 / 28

Recognition. Sanja Fidler CSC420: Intro to Image Understanding 1 / 28 Recognition Topics that we will try to cover: Indexing for fast retrieval (we still owe this one) History of recognition techniques Object classification Bag-of-words Spatial pyramids Neural Networks Object

More information

Subspace Analysis and Optimization for AAM Based Face Alignment

Subspace Analysis and Optimization for AAM Based Face Alignment Subspace Analysis and Optimization for AAM Based Face Alignment Ming Zhao Chun Chen College of Computer Science Zhejiang University Hangzhou, 310027, P.R.China zhaoming1999@zju.edu.cn Stan Z. Li Microsoft

More information

Image Classification for Dogs and Cats

Image Classification for Dogs and Cats Image Classification for Dogs and Cats Bang Liu, Yan Liu Department of Electrical and Computer Engineering {bang3,yan10}@ualberta.ca Kai Zhou Department of Computing Science kzhou3@ualberta.ca Abstract

More information

Feature Tracking and Optical Flow

Feature Tracking and Optical Flow 02/09/12 Feature Tracking and Optical Flow Computer Vision CS 543 / ECE 549 University of Illinois Derek Hoiem Many slides adapted from Lana Lazebnik, Silvio Saverse, who in turn adapted slides from Steve

More information

ColorCrack: Identifying Cracks in Glass

ColorCrack: Identifying Cracks in Glass ColorCrack: Identifying Cracks in Glass James Max Kanter Massachusetts Institute of Technology 77 Massachusetts Ave Cambridge, MA 02139 kanter@mit.edu Figure 1: ColorCrack automatically identifies cracks

More information

Learning Visual Attributes

Learning Visual Attributes Learning Visual Attributes Vittorio Ferrari University of Oxford (UK) Andrew Zisserman University of Oxford (UK) Abstract We present a probabilistic generative model of visual attributes, together with

More information

Geometric Camera Parameters

Geometric Camera Parameters Geometric Camera Parameters What assumptions have we made so far? -All equations we have derived for far are written in the camera reference frames. -These equations are valid only when: () all distances

More information

Data, Measurements, Features

Data, Measurements, Features Data, Measurements, Features Middle East Technical University Dep. of Computer Engineering 2009 compiled by V. Atalay What do you think of when someone says Data? We might abstract the idea that data are

More information

Image Stitching using Harris Feature Detection and Random Sampling

Image Stitching using Harris Feature Detection and Random Sampling Image Stitching using Harris Feature Detection and Random Sampling Rupali Chandratre Research Scholar, Department of Computer Science, Government College of Engineering, Aurangabad, India. ABSTRACT In

More information

The calibration problem was discussed in details during lecture 3.

The calibration problem was discussed in details during lecture 3. 1 2 The calibration problem was discussed in details during lecture 3. 3 Once the camera is calibrated (intrinsics are known) and the transformation from the world reference system to the camera reference

More information

Point Matching as a Classification Problem for Fast and Robust Object Pose Estimation

Point Matching as a Classification Problem for Fast and Robust Object Pose Estimation Point Matching as a Classification Problem for Fast and Robust Object Pose Estimation Vincent Lepetit Julien Pilet Pascal Fua Computer Vision Laboratory Swiss Federal Institute of Technology (EPFL) 1015

More information

A Bayesian Framework for Unsupervised One-Shot Learning of Object Categories

A Bayesian Framework for Unsupervised One-Shot Learning of Object Categories A Bayesian Framework for Unsupervised One-Shot Learning of Object Categories Li Fei-Fei 1 Rob Fergus 1,2 Pietro Perona 1 1. California Institute of Technology 2. Oxford University Single Object Object

More information